Coursera Course Scraper

Coursera's course catalogue, with two numbers most catalogue scrapers leave out: how many people enrolled and how many viewed the page last month, plus the learner reviews. 57 fields per course, as JSON, CSV or Excel.

Five things that are different here

1. Enrollment and traffic, not just a catalogue listing. totalEnrollmentCount and pageViewsInLastMonthCount come straight off the course page. Andrew Ng's machine learning course: 1,248,467 enrolled, 1,718,986 page views in the last month. That is how big the course actually is, not just what it is called.

2. The full rating split, not an average. A 4.89 from 32,913 ratings is not the same thing as a 4.89 from nine. You get the count for every star level, the separate instructor rating and its count, and Coursera's own content satisfaction score with its sample size.

3. The syllabus, lecture by lecture. Module names with their total duration, how many lectures and how many assessments in each, the total course minutes, and whether any item is AI-graded. That is course structure you can compare across providers, not a marketing blurb.

4. Reviews, with the star and the date attached. Up to 20 learner reviews per course: the text, the star rating, the date and the two-letter initials Coursera itself displays. An aggregate of 4.89 tells you people liked it; the reviews tell you what they liked, and the dates tell you whether the praise is current. Measured across 43 courses: 304 reviews, median 18 where a course has any. Turn them off with one switch if you only want the numbers.

5. It walks Coursera's own sitemap. Coursera's robots.txt closes every search and query URL, and the catalogue pages only ever return the same twelve featured courses. So this Actor enumerates from the sitemap Coursera publishes itself. Measured on 18 September 2026: 21,985 course URLs in one pass.

What you get

Identity Course id, name, slug, URL, product type, course type, status, difficulty level, launch date.

Description Full description, estimated workload in the provider's own words, learning objectives and recommended background as clean text.

Provider and instructors Partner names and count. Per instructor: name, title, department, how many courses they teach and how many learners they have reached. Plus the top-instructor count and the total learners reached across all of them.

Reach and reception Enrollment count, page views in the last month, rating average and count, five separate star counts, instructor rating and count, content satisfaction score and its sample size.

Reviews Up to 20 per course, each with star rating, review text, review date and the initials Coursera displays. Plus a review count, the average of the sampled stars and the most recent review date, so you can sort by freshness.

Structure Module count, per-module duration and lecture and assessment counts, total lectures, total assessments, total material items, total minutes, and an AI-grading flag.

Skills and languages Skill tags with a count, the subset flagged as tools or software, primary languages, and how many subtitle, translated and dubbed languages exist.

Provenance sourceUrl, fetchedAt, httpStatus and filledFieldCount travel with every record.

Typical uses

Sizing a topic before building a course of your own: enrollment and monthly views tell you where demand actually is. Benchmarking your catalogue against a competing provider. Finding which skills the market is teaching and at what depth. Tracking a provider's reach over time by running on a schedule.

Example record (trimmed)

{
  "name": "Advanced Portfolio Construction and Analysis with Python",
  "partnerNames": ["EDHEC Business School"],
  "difficultyLevel": "INTERMEDIATE",
  "totalEnrollmentCount": 26411,
  "pageViewsInLastMonthCount": 10958,
  "ratingAverage": 4.745,
  "ratingCount": 514,
  "fiveStarCount": 392,
  "oneStarCount": 6,
  "contentSatisfactionScore": 96.4,
  "reviewCount": 20,
  "reviewSampleAverage": 4.85,
  "latestReviewAt": "2026-04-13T00:00:00.000Z",
  "reviews": [
    { "rating": 5, "authorInitials": "AC", "reviewedAt": "2024-10-08T00:00:00.000Z",
      "comment": "great course. Nice explanation even if you are not familiar with certain mathematical concepts" }
  ],
  "moduleCount": 4,
  "lectureCount": 38,
  "totalDurationMinutes": 733,
  "containsAiGrading": false,
  "skillCount": 14,
  "subtitleLanguageCount": 12,
  "filledFieldCount": 48,
  "courseUrl": "https://www.coursera.org/learn/advanced-portfolio-construction-python"
}

Input

Pick which parts of the sitemap to walk: courses, specializations, professional certificates, guided projects. Add urlKeywords to keep only URLs containing a word such as python. Add exact course URLs if you already know them.

includeReviews is on by default and reviewLimit caps how many reviews per course are kept, up to the 20 Coursera exposes.

maxItems caps how many records are written so a long run cannot produce an unexpected charge.

Failures are loud

The run fails with an explanation rather than finishing quietly. If the page structure changes, you get a parse error naming the affected pages, not an empty dataset. Counters are reconciled every run: sitemap requests, URLs found, filtered out, pages fetched, courses written, duplicates, robots-blocked, HTTP failures, parse failures. The breakdown goes to RUN_STATS, every skipped page to SKIPPED.

Limits and compliance

robots.txt is checked before every request and a disallowed path is never fetched. Coursera closes /search and every ?query= URL, so this Actor never attempts them; when one is supplied it is reported as ROBOTS_DISALLOW and skipped. If robots.txt itself cannot be read, the run stops before making a single data request.

Reviews carry no learner identifier. Coursera publishes review authors as two initials only, never a full name; measured on 18 September 2026 across 120 reviews, 120 of 120 were initials. What Coursera does attach is a numeric learner id inside the review key, and that id is stable across courses, so it is the one field that would let anyone follow a person around the catalogue. It is dropped before the record is written. You get the rating, the text, the date and the initials as shown on the public page. Set includeReviews to false to leave reviews out entirely.

Instructors appear in their professional capacity, as the authors of the product: name, title, department and teaching statistics. No photos, no personal contact details.

Pricing

$2.50 per 1,000 courses, everything included. No start fee, no minimum charge per run, no separate platform usage line.

Paid Apify plans pay less, automatically — there is nothing to apply for: Bronze $2.25, Silver $2.00, Gold and above $1.75 per 1,000 courses. The Free plan price is unchanged at $2.50.

You are charged per course written. Pages that produce nothing, duplicates and pages blocked by robots.txt are reported in the run statistics and cost you nothing.

Measured performance

Run on 18 September 2026 from Apify's own network, no proxy:

Course URLs found in the sitemap	21,985
Pages fetched	43
Records written	43
Duplicates	0
Parse failures	0
HTTP failures	0
Fields populated per course	47 of 57 (median)
Reviews collected	304
Courses carrying reviews	22 of 43
Reviews per course, where present	18 (median), 20 (max)

The 21 courses that returned no reviews have a median rating count of zero — Coursera holds no reviews for them, so there is nothing to miss. Courses that do carry reviews average 27,927 enrollments; the ones that do not average 501. Review text runs to 199 characters, which is the length Coursera publishes on the course page.